Hmm not sure if it means it can self-improve itself exactly. It feels like a lot of variables involved there and often the code produced is just an approximatio

Hmm not sure if it means it can self-improve itself exactly. It feels like a lot of variables involved there and often the code produced is just an approximation of an existing requirement / something in the training data. But yeah, I’m not quite sure what the end goal of l this research is supposed to be either. Seems like we’re not acting very intelligently on many fronts.